The Bay Area Travel Writers (BATW) acknowledges the potential of AI tools for research, reduced busywork and as a potentially “collaborative” tool for human writers and creators. We encourage members to familiarize themselves with their use and where they might facilitate individual efforts. This is a relatively young technology with uncharted possibilities.
Unfortunately, the novelty of these tools has obscured the reality that they are frequently built upon countless involuntary contributions from decades of original producers who were not compensated for their efforts and did not sign releases to their work.
As a result, the BATW maintains that AI-generated content should not be admitted into any work members wish to publish as completely their own. This includes instances in which a member wishes to acknowledge the use of AI-generated content, but is unable to accurately credit the creators whose work powered the AI-generation.
Until a time when common generative AI technologies paired with ethically-sourced training material are the industry default, we encourage members to cautiously explore means of using these tools while maintaining the journalistic standards they uphold elsewhere. In particular, this means deeply researching source materials used when training AI models that a member is considering using as well as transparency regarding their use in works.
BATW stance on text generators
E.g. ChatGPT, Bing Chat, NotionAI, GrammarlyGO
We will not publish, nor endorse the publication of, articles which are partially or fully generated by AI unless it is conditional to the point of the story and disclosed before the generated content. This includes providing background for portions of a piece in which the writer is not an expert — it is your duty as a writer to find an expert — as well as “polishing” or “tone alteration” in which a writer has asked a text generator tool to improve upon a text beyond simple spellchecking and grammar proofing. This includes the abilities of professional-grade software for which a member may have paid.
BATW stance on image generators
E.g. MidJourney, Dall-E, Adobe Firefly
We will not publish, nor endorse the publication of, images which are partially or fully generated by AI unless it is conditional to the point of the story and disclosed before the generated content. This includes images which you are unable to produce — it is your duty as a creator to find photography you can use or write in a way that does not require it — as well as “photoshopping” in which a creator has asked an image generator to enhance, modify or otherwise edit an image which originally belonged to them beyond simple touchup. This includes the abilities of professional-grade software for which a member may have paid.
BATW stance on usage of AI tools outside of content creation
We will publish and endorse the usage of AI tools in preliminary research for an article to the extent the tool is used externally of the submitted work. This may include asking ChatGPT questions like “What are the top 10 things to do in Monterey Bay?” as a starting place for your own research; we believe this to be little different in results from asking a non-AI search engine the same query. This is equally true of more nuanced searches in which AI assistants tend to excel, e.g., “Tell me if there are any Michelin-starred chefs in California who have also worked in France.” There are, however, still ethical considerations concerning from where this information came and whether its creators intended for it to be presented as such, even if you are not directly using the material.
As such, we expect and encourage our members to continue prioritizing their own experiences — or that of locals — to whatever results such searches might produce, just as they might dig deeper than the listicles produced by any search engine. Our audiences enjoy our work precisely because they are not able to visit and experience destinations the same way we can; relying on AI reduces that value.
We also expect our members to recognize the established tendency of search engines— both pre and post generative AI — to “hallucinate” information. Large language models, upon which most generative text AIs are built, aer in particular susceptible to passing off entirely fabricated information as fact, and much like unverified sources elsewhere on the internet, must be fact-checked.
Finally, AI tools are useful in providing editorial feedback and may be used as such when preparing a piece for submission to an editor, so long as the text generator is not performing the work for you (e.g., “make this article more sophisticated”). Use cases in which the generator is working entirely with your own material are also acceptable (e.g., “condense this piece to 300 words”), but it is our hope you’ll rely on your own writerly skills — or that of a human counterpart, such as your fellow members — to provide genuine feedback rather than what a generator algorithmically determines is correct.